biorxiv-search

name	biorxiv-search
description	Fetch bioRxiv and medRxiv preprint metadata by DOI, or browse a narrow date range with a category, for life-science and medical preprints. Use when you already have a DOI or an approximate date plus category — not for open-ended keyword discovery. Public API, no credential needed.
license	Apache-2.0

bioRxiv / medRxiv Search (Codex-native)

Codex-native adaptation of Google DeepMind's literature-search-biorxiv skill (science-skills, Apache-2.0). Uses the public bioRxiv API directly.

Read first: this is a date archive, not a search engine

The bioRxiv API has no server-side keyword search. Choose by what you know:

• DOI (from a citation): fetch metadata directly. Fast and reliable.
• Approximate date + category: browse a narrow (1–4 week) range with a category.
• Only a topic/keywords: do not discover here. Use a keyword-capable skill first ($cx-pubmed-search, $cx-europepmc-search, $cx-openalex-search, $cx-arxiv-search) to find DOIs, then return here for preprint metadata.

Anti-pattern: never scan months/years and filter locally by keyword — it triggers thousands of calls, timeouts, and API blocks.

Gates (ask before proceeding)

• Terms notice: on first use, tell the user to review the bioRxiv terms (https://api.biorxiv.org/, https://www.biorxiv.org/content/about-biorxiv) and to check each retrieved paper's own license; record the notice in provenance.
• Network / install: read-only public HTTP; be polite about rate limits. If you write helper code, run it via uv run, never bare python3.

Workflow

1. By DOI: https://api.biorxiv.org/details/biorxiv/<DOI> (use medrxiv for medRxiv). Returns title, authors, date, category, version, and abstract.
2. By date + category: https://api.biorxiv.org/details/biorxiv/<YYYY-MM-DD>/<YYYY-MM-DD>/<cursor>; page with the cursor and filter by category over a narrow window. Write responses to a file and parse them.
3. Report each preprint with title, DOI, date, category, and URL; note the version.
4. Provenance & review: record DOIs/queries with $science-provenance; check claims with $science-review.

Boundaries

• Preprints are not peer-reviewed; flag that and prefer the published version when one exists (check for a published-DOI link).
• Cite only DOIs actually returned; never invent DOIs or abstracts.
• Respect each preprint's individual license before reproducing content.